Semantic segmentation and quantitative analysis of tunnel cracks and water leakage using a TransUNet framework
Xinjian Li, Qiaofeng Liu, Gang Yan, Qifeng Yu · 2026
Read Semantic segmentation and quantitative analysis of tunnel cracks and water leakage using a TransUNet framework by Xinjian Li, Qiaofeng Liu, Gang Yan, Qifeng Yu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0349175.txt.
Description
Xinjian Li and co-authors trained a hybrid vision model to outline cracks and water leaks on tunnel linings. Inspectors miss damage under poor light and rough surfaces, and simple threshold software fails there too. Their 2026 framework, TransUNet, marries convolutional networks, which read local texture, with transformers, which track long shapes across an image.
The training set mixed public benchmarks with real engineering photographs, all prepared through one pipeline. Against leading rivals, the hybrid scored best on crack outlining, reaching a solid overlap score. Thin, blurred defects and boundary detail improved most. Li and colleagues propose the system for routine tunnel health monitoring, where early cracks and leaks are cheapest to fix.
Catalog record
| CREATOR | Xinjian Li, Qiaofeng Liu, Gang Yan, Qifeng Yu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50031 |
| LICENSE | CC_BY |
| RIGHTS | CC BY |
| SOURCE | PLOS |